Google releases Gemini Robotics 2.5, a vision-language-action model with 96% success on Mercedes-Benz lines

Google says Gemini Robotics 2.5 is now rolling out to its robotics partners. The company reports a 30% to 100% higher success rate than previous baselines depending on task type, with tasks completed 20% to 30% faster, evaluated across 200 distinct tasks such as serving food with tongs and pipetting liquid. In a real-world proof-of-concept, the model helped Apptronik's Apollo 2 humanoid pick and pack automotive parts on Mercedes-Benz production lines at a 96% success rate. Google also introduced a benchmark testing adversarial jailbreaks specific to robotics and says Gemini 4 Argon accelerated development by acting as a simulation engine.

✨ AI summary of the linked source — not editorial content.

Explore the topic

• Google
• Gemini